Open

Blog

Tamara News  /  Blog
04 Sep 2026
Data center servers behind the AI chatbot outage

ChatGPT, Claude and Grok All Went Dark at Once — Here’s Why

Categories Tech

An AI chatbot outage knocked three of the world’s most-used AI tools offline within the same hour. It happened on September 3, 2026. OpenAI’s ChatGPT, Anthropic’s Claude and xAI’s Grok all began throwing errors around 7:43 a.m. Pacific time. Downdetector logged sharp spikes in error reports across all three services within minutes of each other.

Three competing AI labs rarely fail at once. Chatbots from different companies almost never share a single point of failure. The near-simultaneous timing pointed straight at the infrastructure underneath them, not at any one model.

How the AI Chatbot Outage Unfolded

ChatGPT’s disruption did not stop at the chat window. OpenAI confirmed that logins, file uploads, voice mode, search, Deep Research and image generation were all affected. Its Codex coding tool went down too, according to 9to5Mac. Anthropic said its newest Claude Opus 4.8 and Opus 5 models took the hardest hit. Older Claude models recovered to normal error rates faster. xAI’s Grok also went down, and the company acknowledged the problem publicly while it worked on a fix.

Secondary services built on top of these models felt it too. Developer tools including the coding agent Cursor confirmed downtime, since they depend on the same upstream AI infrastructure. The Register first reported that knock-on effect. All three chatbots returned to normal service by 12:38 p.m. Pacific time.

What Likely Caused the AI Chatbot Outage

Microsoft Azure was also experiencing its own outage at the same time. Several outlets, including Axios, said Microsoft’s cloud problems may have contributed to the AI outages. None of the three AI companies has published a full technical explanation yet. Tamara News has not independently verified a single root cause. This is what multiple outlets reported as the likely trigger, not a confirmed finding from OpenAI, Anthropic or xAI themselves.

The pattern fits a broader worry in the industry. Even rival AI companies often lean on the same handful of cloud providers, data center regions and networking backbones. When one shared layer stumbles, the outage rarely stays contained to a single brand.

This is not the first time a cloud provider’s problems have rippled outward into services that look unrelated on the surface. Streaming platforms, banking apps and airline booking systems have all gone dark in past incidents traced back to a single cloud region. AI chatbots are simply the newest category to join that list, and arguably the most visible one, since millions of people now open ChatGPT the way they once opened a search engine.

Why This AI Chatbot Outage Matters Beyond the Inconvenience

For casual users, an hour without ChatGPT is an annoyance. For businesses that wired these models into customer support, coding pipelines and internal workflows, it becomes a production incident. Companies that built single-vendor dependencies into critical systems had no working fallback during the window. The model they might have switched to was also down.

This is the second major cloud-linked disruption Tamara News has covered in three days. A separate Microsoft cloud outage hit a different set of services on September 2. Together, the two incidents raise the same question from different angles: how much of daily digital life now depends on a small number of cloud platforms staying up.

Enterprise customers with strict uptime requirements are the ones most exposed here. A five-hour gap in service can cascade into missed customer replies, stalled code deploys and delayed research work. Few companies have built real fallback plans for a scenario where every major AI vendor drops at once.

Some engineering teams responded by switching to open-weight models they can run on their own hardware, at least temporarily. Others simply waited it out. Either response points to the same underlying gap: most organizations that rely on AI chatbots day to day have not stress-tested what happens when none of them are reachable.

What Happens in the Days Ahead

Expect each company to publish, or decline to publish, a post-incident report in the coming days. OpenAI, Anthropic and xAI have varying track records on transparency after outages. Some post detailed timelines. Others post a short status update and move on. Enterprise customers with service-level agreements will likely ask pointed questions about compensation and redundancy planning. None of the three companies has said whether Wednesday’s incident will change how they architect around single points of failure.

AI chatbot outage error screen on a smartphone

Frequently Asked Questions

What caused the AI chatbot outage on September 3, 2026?
No company has issued a confirmed root-cause statement. Multiple outlets, including Axios, reported that a concurrent Microsoft Azure disruption may have contributed. OpenAI, Anthropic and xAI have not confirmed this independently.

How long did ChatGPT, Claude and Grok stay down?
Reports place the start of the disruption around 7:43 a.m. Pacific time. All three services returned to normal by roughly 12:38 p.m. Pacific — just under five hours.

Were all ChatGPT features affected?
Yes. OpenAI confirmed that logins, uploads, voice mode, search, Deep Research, image generation and its Codex coding tool were all disrupted, not just the main chat interface.

Did the outage affect tools built on top of these chatbots?
Yes. Developer tools including the coding agent Cursor reported downtime because they depend on the same upstream AI infrastructure.

Is this connected to the Microsoft cloud outage from September 2?
The two incidents happened a day apart and both involve cloud infrastructure. Tamara News has not seen a confirmed technical link between them.

Featured image: “BalticServers data center.jpg” by BalticServers.com (CC BY-SA 3.0), via Wikimedia Commons. In-content image: “Smartphone with ChatGPT app (52917381673).jpg” by Jernej Furman from Slovenia (CC BY 2.0), via Wikimedia Commons.

Related Guides

02 Sep 2026
US travel document representing cases affected by the US immigrant visa freeze

Wondering Why Your Green Card Case Has Stalled? This Policy Might Be Why

Categories Immigration, Visa Updates

The US immigrant visa freeze that paused green card processing for nationals of 75 countries has now been in effect for more than seven months. The State Department has given no announced end date. Applicants and families still asking why their case has stalled are often running into this policy rather than routine backlog delays.

The freeze took effect January 21, 2026, a week after the State Department announced it. It applies only to immigrant visas, the category that leads to a green card, not to tourist visas or most other nonimmigrant categories.

What the US immigrant visa freeze actually covers

PBS NewsHour reported the State Department cited concerns about applicants relying on public benefits as the stated reason for the pause. The freeze reaches family members of U.S. citizens, including spouses and children abroad. It also covers immigrant workers already in the process of employer sponsorship, provided they hold a passport from one of the 75 listed countries.

US travel document representing cases affected by the US immigrant visa freeze

Photo: U.S. Citizenship and Immigration Services, public domain, via Wikimedia Commons.

Who is exempt from the freeze

The Council on Foreign Relations notes that this 75-country list is separate from, and broader than, the roughly 39-country entry travel ban. That narrower ban restricts actual admission to the United States. Countries such as Brazil, Egypt, Iraq, Jordan, and Thailand appear on the visa-freeze list without being subject to the narrower travel ban. Dual nationals holding a valid passport from a country not on the list are exempt from the freeze. That holds even if they were born in an affected country.

Travel document paperwork tied to cases affected by the US immigrant visa freeze

Photo: public domain, via Wikimedia Commons.

How families are affected in practice

Immigration attorneys report that cases already scheduled for a final immigrant visa interview before January 21 mostly proceeded on schedule. New interview scheduling for applicants from the 75 countries has largely stopped since then. Families midway through consular processing describe months of silence from the National Visa Center. There is no published mechanism for requesting a case-by-case exception outside of narrow humanitarian circumstances.

What legal challenges could change

Several immigrant advocacy organizations have signaled they are preparing lawsuits arguing the freeze exceeds the administration’s authority under existing immigration law. No court has yet issued an order blocking or narrowing the policy. Attorneys tracking the litigation say a ruling could take months. Similar immigration cases have moved through federal appeals courts at varying speeds over the past two years. Until then, families affected by the freeze have no clear timeline for when normal processing might resume.

Previous administrations have used narrower versions of visa pauses tied to specific countries during public health or security emergencies. A blanket freeze covering 75 countries at once has no direct recent precedent, according to immigration historians. Some members of Congress have introduced bills on the freeze. Those bills would require the State Department to justify it on a country-by-country basis rather than as a single blanket policy. None of those bills has advanced out of committee. Supporters of the freeze argue that a case-by-case review would defeat the administrative efficiency the policy is meant to provide.

What happens while the freeze stays in place

With no announced expiration date, immigration lawyers advise affected clients to keep their paperwork current. They also recommend continuing to pay required fees so a case does not lapse, even though processing itself has stalled. Advocacy groups have signaled possible legal challenges, though none have yet produced a court order affecting the freeze’s operation. For related coverage on other active immigration changes, see our reporting on the USCIS public charge rule update and the elimination of Duration of Status for student visas.

The public-charge rationale behind the freeze echoes similar restrictions the administration has applied to other visa categories this year. That includes tightened scrutiny of certain nonimmigrant applications. Immigration policy researchers note that public-charge related restrictions have expanded and contracted across different administrations for decades. That history makes them a frequent target of legal challenges. Families affected by the current freeze say the lack of a published review timeline is the hardest part. Even a temporary policy can permanently derail plans tied to jobs, housing, or school enrollment abroad.

Freeze FAQ: what’s still true

When did the US immigrant visa freeze start?
The freeze took effect January 21, 2026, following a State Department announcement made a week earlier.

Which visas does the freeze affect?
Only immigrant visas, the category that leads to a green card. Tourist visas and most other nonimmigrant visas are not affected.

How many countries are on the list?
75 countries, including Brazil, Egypt, Iran, Russia, and Somalia among others.

Is this the same as the travel ban?
No. The 75-country visa freeze is broader than, and separate from, the roughly 39-country entry travel ban.

Are any applicants exempt?
Dual nationals with a valid passport from a country not on the list are generally exempt, even if born in an affected country.

Is there an end date for the freeze?
No end date has been announced, and the policy remains in effect as of this writing.

Immigration attorneys say the freeze has increased demand for consultations from families trying to understand whether their specific case qualifies for an exemption. Some law firms have created dedicated intake forms just to handle the volume of questions tied to the 75-country list. Community organizations serving affected populations say they are focusing outreach on explaining the difference between the visa freeze and the separate travel ban. The two are often confused.

Sources:

02 Sep 2026
US Embassy building in Africa affected by the US visa services Africa hub realignment

The US Just Closed Routine Visa Services at 25 African Embassies — Here’s Where to Go Instead

Categories Immigration, Visa Updates

US visa services Africa changes took effect August 1, 2026. Routine visa processing shifted away from 25 individual embassies and consulates into 20 regional hubs across the continent. The State Department says the move standardizes screening and adjudication rather than closing any diplomatic post.

Applicants in affected countries must now book appointments and pay fees at a designated regional hub rather than their home-country embassy. The change touches both nonimmigrant visas, such as tourist and student visas, and immigrant visas that lead to permanent residency.

How the US visa services Africa realignment works

According to the State Department, routine visa services in several cities have moved to a nearby regional hub. Affected posts include Abuja, Bamako, Banjul, Conakry, Freetown, Harare, Juba, Lusaka, Maputo, and roughly fifteen others. The Department frames the shift as consistent with long-standing practice elsewhere in the world. Officials say it should improve consistency in vetting standards and speed up processing at higher-volume locations.

US Embassy building in Africa affected by the US visa services Africa hub realignment

Photo: U.S. Embassy Nairobi, public domain, via Wikimedia Commons.

What applicants need to do differently

Fragomen, an immigration law firm tracking the change, notes that affected posts remain open for limited or emergency consular services. Applicants seeking a new visa must travel to the designated hub for their region to complete interviews and biometrics. That can mean additional travel time and cost for some applicants. Those in landlocked or remote countries may find their nearest hub sits in a different nation entirely.

Passport control counter illustrating the visa processing changes under US visa services Africa realignment

Photo: Manila International Airport Authority, public domain, via Wikimedia Commons.

Why the State Department made this change

Officials describe the hub model as a security and efficiency measure. It lets a smaller number of specialized consular teams handle higher visa volumes with more uniform training and oversight. That beats spreading staff thinly across dozens of smaller posts, officials argue. Critics, including some regional business groups, warn the change could slow down travel and trade-related visa applications in the short term. Applicants and travel agents still need to adjust to new booking systems and locations.

How travel and education sectors are adjusting

Study-abroad advisers say prospective students now need to budget extra weeks for visa appointments if their nearest hub sits outside their home country. Airlines serving smaller African markets report early signs of shifting booking patterns. More connecting flights now route through hub cities ahead of scheduled visa interviews. Tour operators handling group travel to the United States say they have started building the realignment into standard trip-planning timelines. They no longer treat it as a temporary disruption.

Some African governments have privately raised concerns about the realignment through diplomatic channels. That is according to immigration lawyers who work with clients in the affected countries. Officials in several capitals say they were given limited advance notice before the August 1 effective date. Business associations that rely on frequent US visa travel for trade delegations want faster options. They are lobbying for expedited appointment slots at the new hub locations. The State Department has not announced any plans for an expedited track specific to business travelers affected by the consolidation.

What happens next for applicants

The State Department has not signaled any plan to reverse the realignment. Applicants in the affected countries should expect the hub system to remain the standard going forward. Travel agents and visa service companies operating in the region are updating guidance to help clients identify their new designated hub and book appointments accordingly. For related coverage on shifting visa rules elsewhere, see our reporting on the September 2026 visa bulletin retrogression and the extended Schengen border checks.

The State Department has used regional hub models in other parts of the world for years, including in parts of the Caribbean and Central Asia. Officials argue the approach concentrates experienced staff where visa demand is highest. Immigration lawyers say the tradeoff is consistent: applicants gain more predictable processing standards but lose the convenience of a nearby post for routine paperwork. Some regional hubs have already reported longer appointment wait times as they absorb caseloads from the newly consolidated posts. The State Department has not published hub-by-hub wait time data.

Visa hub questions, answered

When did the US visa services Africa change take effect?
The realignment took effect August 1, 2026.

How many posts and hubs are involved?
Routine visa services moved from 25 individual posts into 20 regional hubs across the continent.

Does this affect immigrant and nonimmigrant visas equally?
Yes, the realignment applies to both immigrant visas, which lead to permanent residency, and nonimmigrant visas such as tourist and student visas.

Are the affected embassies closing?
No. The State Department says affected posts remain open for limited or emergency consular services, but not routine visa processing.

Will applicants need to travel further?
In many cases, yes, since some applicants must now travel to a hub located in a neighboring country to complete their visa interview.

Is the change permanent?
The State Department has given no indication it plans to reverse the realignment, so applicants should plan around it for the foreseeable future.

Visa consultants say the change has created a small but noticeable market for local agents who help applicants book hub appointments and arrange travel. Some applicants have reported confusion over which hub now serves their country, since the mapping was not always intuitive based on geography alone. The State Department has published a country-by-hub reference list to reduce that confusion going forward.

Sources:

02 Sep 2026
Microsoft office signage pictured during the Microsoft cloud outage September incident

A Forgotten Certificate Just Took Down Microsoft 365 for Millions of Users

Categories Tech

The Microsoft cloud outage September incident started with an authentication certificate problem in Exchange Online. It spread within hours to Teams, SharePoint, OneDrive, and Microsoft Defender XDR, disrupting work for businesses worldwide. Microsoft’s own status page tracked the incident under the internal code EX1464935.

The trouble began around 8 a.m. Pacific time on Monday. By Tuesday, Microsoft said core Exchange Online mail flow and search had recovered, though it acknowledged some services were still not fully back to normal.

What triggered the Microsoft cloud outage September incident

Born’s Tech and Windows World reported the root cause traced back to a core authentication configuration issue tied to a certificate that had not been renewed. TechCrunch confirmed the outage dragged into a second day. Users kept reporting intermittent search and calendar failures even as mail delivery stabilized.

Microsoft office signage pictured during the Microsoft cloud outage September incident

Photo: Shijaz, public domain, via Wikimedia Commons.

Which services and users were affected

BleepingComputer listed symptoms ranging from delayed or failed email delivery to broken mailbox search. Teams users also saw calendar problems, and their presence status got stuck instead of updating automatically. Microsoft 365 Business and Enterprise customers were hit hardest, since those plans lean most heavily on Exchange Online for daily operations. Consumer Outlook.com accounts saw more limited disruption by comparison.

Laptop representing the email and productivity tools disrupted by the Microsoft cloud outage September incident

Photo: Raimond Spekking, CC BY-SA 4.0, via Wikimedia Commons.

How Microsoft has responded so far

Microsoft rotated the expired certificate and pushed a fix across affected regions in stages. That staged rollout explains why recovery came unevenly, with some tenants back to normal well before others. The company has not yet published a full post-incident report explaining why the certificate renewal was missed in the first place. Outage-tracking sites note this is not the first cloud-wide disruption tied to a routine maintenance failure this year.

What it cost businesses on the ground

Help-desk teams at several large enterprises reported a spike in support tickets tied to failed logins and missing calendar invites during the outage window. Small businesses that rely on Microsoft 365 for customer-facing email said delayed replies likely cost them time-sensitive deals. No company has published a dollar estimate of the damage. IT consultants say the incident has renewed client interest in backup email routing and offline access to critical files. Many organizations had let those precautions lapse during years of relatively stable service.

Some enterprise customers in regulated industries, including banking and healthcare, say outages like this one complicate their compliance obligations. Several regulators require documented uptime for systems handling sensitive data. A handful of large customers have asked Microsoft for compensation under existing service-level agreements. The exact terms of those agreements vary by contract and are not publicly disclosed. Cybersecurity researchers have also noted that widescale outages create a brief window of confusion. Phishing campaigns sometimes exploit it, since employees expect unusual login prompts and error messages while a real incident is underway.

What happens if outages like this keep happening

Enterprise customers are increasingly asking Microsoft for clearer service-level commitments. They also want faster public disclosure when an outage begins, rather than learning about scope and cause well after the fact. Industry analysts expect cloud outages to remain a recurring risk as more critical business functions concentrate on a small number of providers. A single configuration failure can now ripple across email, chat, and file storage simultaneously. For related coverage on the accountability pressure facing major platforms, see our reporting on the EU’s ChatGPT search engine designation and the ongoing wave of tech layoffs.

Industry publications have warned throughout 2026 that cloud outages are becoming more frequent. Providers are packing more services onto shared authentication and identity systems. When one of those shared systems fails, the effects cascade across products that otherwise have little in common. That is part of why a single expired certificate could take down email, chat, and file storage at the same time. Competing cloud providers have used past Microsoft outages to court enterprise customers considering multi-cloud strategies. Switching core productivity software still remains a slow and costly process for most large organizations.

Outage FAQ: what to know

What caused the Microsoft cloud outage September incident?
Microsoft traced the root cause to a core authentication configuration issue linked to an expired certificate.

Which services were affected?
Exchange Online, Teams, SharePoint, OneDrive, Microsoft Purview, and Defender XDR all reported disruptions.

How long did the outage last?
The incident began Monday morning and stretched into a second day, with core mail flow restored first and some services lagging behind.

Were consumer accounts affected?
Consumer Outlook.com users saw more limited disruption compared with Microsoft 365 Business and Enterprise customers.

Has Microsoft explained why the certificate lapsed?
Not yet. Microsoft has not published a full post-incident report detailing why the renewal was missed.

Could this kind of outage happen again?
Analysts say concentrated reliance on a handful of cloud providers makes repeat incidents likely, since one configuration failure can affect multiple services at once.

Cloud reliability researchers say the incident adds to a growing case study list used in enterprise risk assessments. Several large customers say they now factor outage history directly into vendor renewal negotiations. Microsoft has not said whether it will offer service credits automatically or require affected customers to file individual claims. That distinction matters for smaller businesses without dedicated legal or procurement teams.

Sources:

02 Sep 2026
Computer processor chip representing the hardware behind the Anthropic Nvidia compute deal

Anthropic Just Locked In $35 Billion of Computing Power — Nvidia Is on Both Sides of It

Categories Business, Tech

The Anthropic Nvidia compute deal is a six-year, $35 billion cloud-computing agreement between the AI lab and Lambda. Nvidia backs Lambda with both equity and hardware. The deal routes through a Texas data center and adds to more than $135 billion in computing contracts Anthropic has signed so far this year.

Under the arrangement, Lambda will draw computing capacity from a 700-megawatt data center campus in Nueces County, Texas. Hut 8, a company known for bitcoin mining and data center development, is building the facility.

How the Anthropic Nvidia compute deal is structured

According to Bloomberg, Nvidia will lease the Hut 8 data center directly. Lambda then pays Nvidia for access to it, and Anthropic in turn pays Lambda for the computing power it needs to train and run its models. Techxplore noted that Nvidia has invested in both companies on either side of the contract. Some analysts have questioned this structure as a form of circular financing that keeps Nvidia’s chips, capital, and revenue moving through the same small set of partners.

Computer processor chip representing the hardware behind the Anthropic Nvidia compute deal

Photo: Intel Free Press, CC BY 2.0, via Wikimedia Commons.

Why Anthropic needs this much computing power

Anthropic has signed a rapid succession of large compute contracts this year to keep pace with rivals training increasingly large models. The company spreads its computing needs across several providers rather than relying on a single cloud vendor. That approach reduces the risk that one supplier’s capacity constraints slow its model development schedule.

Graphics processing hardware similar to the chips underpinning the Anthropic Nvidia compute deal

Photo: Advanced Micro Devices, Attribution, via Wikimedia Commons.

What analysts are questioning about the arrangement

Some market analysts have flagged the layered structure of the deal. Nvidia effectively finances the buildout of a data center it will lease to a company it has invested in. That company then sells access to another company Nvidia has also invested in. Supporters argue this is simply how the AI infrastructure buildout works at scale, since few single companies can finance multi-billion-dollar data center campuses alone. Critics counter that if AI demand ever slows, the interlocking investments could concentrate risk across the same handful of players rather than spreading it.

How rival AI labs are responding

OpenAI and Google have both signed their own large multi-year compute contracts over the past year. Industry watchers describe an accelerating arms race for guaranteed chip access. Cloud providers outside the big three hyperscalers, including Lambda and CoreWeave, have used deals like this one to build credibility. AI labs that want to diversify away from a single vendor now see them as real alternatives. Some venture investors say the scale of these commitments is now a meaningful factor in how they assess an AI lab’s long-term viability.

Total spending commitments across the AI industry for computing infrastructure now run into the hundreds of billions of dollars this year alone. Several financial analysts tracking the sector compiled those figures. Some investors have started asking whether current AI revenue justifies commitments of this size. Many AI labs remain unprofitable even as they sign multi-year contracts worth tens of billions of dollars. Anthropic has not disclosed its current revenue run rate publicly. That makes it difficult for outside analysts to assess how comfortably the company can meet its growing computing obligations.

What happens next for AI compute deals

Anthropic has signaled it will likely sign additional compute agreements before year-end as it prepares to train its next generation of models. The Texas data center itself is still under construction. Practical computing capacity from this specific deal will phase in gradually rather than arriving all at once. For related coverage on the compute race, see our reporting on Salesforce’s stake gain tied to Anthropic and the AI chip interconnect bottleneck affecting the wider industry.

Data centers of this scale draw enormous amounts of electricity. A 700-megawatt campus is large enough to power a mid-sized city. Texas grid operators have flagged rising demand from data center construction as a planning challenge for the years ahead. Local officials in Nueces County say the project is expected to bring construction jobs to the area. Permanent staffing needs for a highly automated data center are typically much smaller than the construction workforce required to build it.

Questions readers are asking about the deal

How much is the Anthropic Nvidia compute deal worth?
The agreement is valued at $35 billion over six years.

Where will the computing power come from?
A 700-megawatt data center campus being built by Hut 8 in Nueces County, Texas.

What role does Nvidia play in the deal?
Nvidia leases the data center from Hut 8 and has invested in both Lambda and Anthropic, the two companies on either side of the contract.

Why are analysts calling this circular financing?
Nvidia’s capital and chips flow through companies it has invested in on both ends of the transaction. Some see that as concentrating risk rather than distributing it.

How much has Anthropic committed to compute overall this year?
This deal adds to more than $135 billion in computing contracts Anthropic has signed so far in 2026.

When will the capacity be available?
The Texas facility is still under construction, so capacity from this deal will come online gradually rather than immediately.

Energy analysts say the growing cluster of large AI data centers in Texas is drawing regulator attention. State officials are tasked with keeping the power grid stable during peak demand. Local water authorities have also asked developers to disclose cooling-system water use, since large data centers can consume significant amounts of water depending on their cooling design. Hut 8 has not yet published detailed environmental impact figures for the Nueces County campus.

Sources:

1 … 15 16 17 18 19 … 174